摘要
针对离场航班延误预测缺少对航路网络结构因素的考虑,以及传统多分类预测难以满足高精度的需求,提出了一种考虑航路网络结构的离场航班延误预测模型。首先,根据离场航班所在终端区的航路网络结构,提出了航路拥挤指标,即航路流量、航路拥挤度和航路网络拥挤度,从航路网络和网络结构2个维度量化分析了拥挤特征,构造了航路拥挤数据集;然后,基于深度神经网络(deep neural network,DNN),构建了考虑航路网络结构的离场航班延误预测模型;最后,分析各类别延误样本比例,调整焦点损失函数的平衡因子以及各模型参数,进行了不同损失函数、不同数据集和不同模型参数的对比实验。结果表明:调整平衡因子后,模型预测准确率提高了2.3个百分点,融入航路拥挤数据集后,准确率继续提高了1.52个百分点,并且最终达到93.47%。可见,本文所提模型能够对离场航班延误作出有效准确判断,为民航相关单位提供决策参考。
Aiming at the lack of consideration of air route network structure factors in departure flight delay prediction,and the difficulty of traditional multi-category prediction to meet the demand of high accuracy,a departure flight delay prediction model considering the air route network structure was proposed.Firstly,according to the air route network structure of the terminal area where the departure flight was located,the air route congestion index was proposed,including air route flow,air route congestion degree,and air route network congestion degree.The congestion characteristics were quantitatively analyzed from the two dimensions of air route network and network structure,and the air route congestion dataset was constructed.Secondly,based on the deep neural network(DNN),a departure flight delay prediction model considering the air route network structure was constructed.Finally,the proportion of delayed samples of each category was analyzed,the balance factor of the focal loss function and the parameters of each model were adjusted,and the experiments with different loss functions,different data sets and different model parameters were comparatively carried out.Experimental results show that after adjusting the balance factor,the prediction accuracy of the model is increased by 2.3 percentage points,and after adding the air route congestion dataset,the accuracy is increased by 1.52 percentage points,and the final accuracy reaches 93.47%.It can be seen that the model is proposed in this paper can make effective and accurate judgments on the departure flight delays,providing decision-making reference for civil aviation related units.
作者
徐海文
汪腾
XU Hai-wen;WANG Teng(College of Science,Civil Aviation Flight University of China,Guanghan 618307,China;College of Civil Aviation Safety Engineering,Civil Aviation Flight University of China,Guanghan 618307,China)
出处
《科学技术与工程》
北大核心
2023年第11期4734-4744,共11页
Science Technology and Engineering
基金
中央高校基本科研业务费专项(J2021-057)。
关键词
航路网络结构
航班延误预测
深度神经网络
拥挤指标
air route network structure
flight delay prediction
deep neural network
congestion index
